Homemade experimental impulse response

[Otto Halmén]

I'll screencast if (or when) this utility turns into something releasable.

In the meantime, here's the next specimen (along with a quick demo), now in proper true stereo: Artificial Hall - test 2

This one doesn't use the ripple tank method. Instead, it calculates the amplitudes and delays of individual reflections using basic trigonometry. It renders a lot faster now. The ripple tank algorithm was getting a bit unwieldy for creating large soundstage-type spaces, which are what I'm primarily interested in making.

[kmlandre]

The more I futz with these, the more I'm finding they make for terrific ER's!

Slap on a bit of Valhalla's "Boston Symphony Hall" and the results are pretty spectacular...

How hard would it be to do a "stage placement" (left-right) version of these? Or am I asking the wrong question?

-- Kurt

[Otto Halmén]

Stage placement is built into the new algorithm, actually. It's just x and y coordinates for a source and one for a pickup in a rectangular room, and then it's just good old Pythagoras for calculating the delay, panning, and amplitude of each reflection until a certain threshold. Each reflection is then rendered as a short, filtered noise burst in the waveform.

I just rendered a test with a front, mid, and a back impulse, but they don't work very well together and the back impulse itself is pretty much unusable. It seems there's an error in the way damping is calculated (or possibly in the way normalization is). I'll have to try analyzing and fixing it another day, as it is getting pretty late where I live.

[kmlandre]

Well,  I'll give you 24 hours!

At that point,  well...I guess I'll probably just continue to wait patiently... ;-)

Excellent work so far, though! 

   -- Kurt

[Otto Halmén]

Thanks, Kurt!

It's a lot of trial and error, but I'll carry on. I'm only a self-taught (and Dad-taught) programmer. I didn't even know what a filter kernel was until a while ago, and I never made anything more complex than a 360° snake game and some kind of 2D fighter anyway. Learning to create audio plugins is still on my todo-list.

I'm hoping to release a small pack of usable impulse responses at the end of this experiment, if not the actual impulse generator itself. IRs generated this way have some nice properties. The ERs, for example, build up gradually into the tail of the reverb, as they are both derived from the same math.

[kmlandre]

Did some more fiddling with this - just love the sound of what you've got!

While I generally agree with Mattias that realism is often overrated, it sure doesn't hurt when things *sound* more real, and I definitely like what this thing does for the ERs...




EDIT - By the way, for those who might care about such things, here's the settings I'm using with the IR (the Otto's Original "Artificial Hall"):





 -- Kurt

[kmlandre]

Stupid gdrive.  I'll have to find another place to upload those screenshots...

 -- Kurt

[Otto Halmén]

TIL that the first algorithm I used is essentially highly unoptimized digital waveguide synthesis in two dimensions and that the second one I used is called the image-source method. Who knew? Well, having a name for these things certainly helps with the learning by Googling.